AMD: Speculation, Rumors, and Discussion (Archive)

[CarstenS]

rectangle != square. (Long) Rectangles in the sense of non-square dies are not uncommon by any means. In fact, most dies that are not memory are not exact squares anyway.

 

[Gipsel]

rectangle != square.

Just to nitpick, the correct notation here would be:
squares ⊂ rectangles
Squares are a proper subset of rectangles, meaning that every square is also a rectangle but not the other way around (there are are of course rectangles which are not squares).

(Long) Rectangles in the sense of non-square dies are not uncommon by any means. In fact, most dies that are not memory are not exact squares anyway.

That is of course true and I have mentioned some of the reasons why we have seen dies with aspect ratios higher than 1 in the past and we will continue to see them. Btw., there also plenty of non-square memory dies.

 

[Silent_Buddha]

Yeah I was like
But I guess most esports are older/less 3D intensive + tend to be played on min-quality for max responsiveness.

More that they are less demanding. DOTA 2, StarCraft 2 (more CPU than GPU demanding), LoL, etc. The more graphically demanding a game, the less people can run it well, the less chance it has of becoming a major eSport game.

Regards,
SB

 

[jacozz]

Here you go:

 

[hoom]

So how big is the GDDR chip?

 

[CSI PC]

Peak vs average, if we consider ATOS peak for AMD because its compute heavy has async and doesn't hurt AMD with tessaltion and thus can get high occupation and utilization on GCN 1.X. In that case the raw throughput rate of hawaii looks really good. But we know AMD have lots of changes coming to frontend and shaders. We also have a few snipits like this so if Polaris average and peak is much closer then GCN1.X then both situations can be true. Then there is the overclock possibilities if you dont care about TDP.

My point was about 390x vs 480 where we already have a real world value from AMD in their live event when they compared it to a 1080FE and so we can work back from that.
Cheers

 

[CSI PC]

Multi GPU probably scales worse on newer cards before drivers are fully there.

A 1.83 scaling in a game is good.
But the context is comparing the 480 to the 390x where we can use the same number as a theoretical comparable (I really doubt the 390x scales that well) to see really how well the 480 does compare to previous cards.
In other words scaling this back to a single card it is weaker and noteably so than a 390x in AoTS, and I am using various reviews to get the single GPU number for the 390x.

Hence why it is a headscratcher why WCCF is saying better sell your 390x and 980....
At best it may match it or be just a little ahead at times (depending upon game but as I said in those situations where the 480 is stronger than the 390x it is very likely the 980 is also stronger).
Cheers

 

[Grall]

$250 for an 8 gig card with 980ti performance for vr is a huge deal.

480 needs to be clearly faster than 390X for me to justify a buy, anything else is pretty fail. It draws less power? Ok, well, my setup handles 390X-level heat; not an issue.

However, less heat, smaller card, AND faster? Yeah, now I'm interested. For $250-equivalent: REALLY interested, because 390X is very hot, and really noisy at full load, even with a good 3rd party cooler instead of the crappy reference design hairblower AMD designed for that gen of cards.

 

[itsmydamnation]

My point was about 390x vs 480 where we already have a real world value from AMD in their live event when they compared it to a 1080FE and so we can work back from that.
Cheers

You have a single data point for a single work load. From where i stand the world looks pretty flat to me........................

 

[CSI PC]

You have a single data point for a single work load. From where i stand the world looks pretty flat to me........................

Which is more than anything else to date that are pure rumours and speculation and is based upon a workload created for the event by AMD to be ideal for 480
But to re-iterate this is a data point on a game perfect for AMD, and about comparing their own architecture to each other with relative performance.
So it is pretty appropriate when articles pop up about a 480 having performance beyond the expectations we have seen with AoTS.
Anway for a 480 to make up around 12-15% in other games is possible (and will happen IMO) it probably needs to be in areas previous GCN were weak, but to reiterate in those situations we will also likely see the Nvidia cards also improving as those games play to their strengths rather than weaknesses.
Context is regarding the WCCF comment saying better sell your 390x and 980 with leak they are releasing today/tomorrow.

Cheers

 

[Deleted member 2197]

We are assuming that these are stock results (but bear in mind that it is PC Partner sample, which might be slightly factory-overclocked). We are working on obtaining overclocking scores, but it may take a while. Of course these are 100% legit scores, unlike some strange charts you saw on Chinese forums. Anyway, this puts RX 480 in between R9 390X and GTX 980. There are no Fire Strike Performance Preset scores yet (which could alter overall % figures a bit).

http://videocardz.com/61005/new-amd-radeon-rx-480-3dmark-benchmarks

 

[eastmen]

480 needs to be clearly faster than 390X for me to justify a buy, anything else is pretty fail. It draws less power? Ok, well, my setup handles 390X-level heat; not an issue.

However, less heat, smaller card, AND faster? Yeah, now I'm interested. For $250-equivalent: REALLY interested, because 390X is very hot, and really noisy at full load, even with a good 3rd party cooler instead of the crappy reference design hairblower AMD designed for that gen of cards.

If its better than the 390x and I can plug in two of them for the same power and tdp usage it will be an okay upgrade.

Like I said I would only buy it if I can crossfire it with the 290. However for my gf who has a 6950 this will be an amazing upgrade since she can now use my rift and we don't have to worry about her power supply and her cooling constraints

 

[CSI PC]

480 needs to be clearly faster than 390X for me to justify a buy, anything else is pretty fail. It draws less power? Ok, well, my setup handles 390X-level heat; not an issue.

However, less heat, smaller card, AND faster? Yeah, now I'm interested. For $250-equivalent: REALLY interested, because 390X is very hot, and really noisy at full load, even with a good 3rd party cooler instead of the crappy reference design hairblower AMD designed for that gen of cards.

Are you using a custom AIB 390x or reference?
I thought the custom ones seemed pretty good, albeit usually 2.5 width for better cooling and noise performance.
Cheers

 

[SimBy]

Are you using a custom AIB 390x or reference?
I thought the custom ones seemed pretty good, albeit usually 2.5 width for better cooling and noise performance.
Cheers

I don't think there ever was a reference 390X.

 

[CSI PC]

SO probably like I mentioned in the other thread with the example of AotS.
There will be times the 390x is going to be faster, and games where traditionally GCN was weaker the 480 has performance just above that of the 390x.
But as they say it is a partner card that may have slight/some OC.
Cheers

 

[CSI PC]

I don't think there ever was a reference 390X.

Ah ok I thought there was still (originally before dropped-replaced completely by custom AIB) a blower version, and I know not all of them are 2.5 width.
Cheers

 

[Kaotik]

SO probably like I mentioned in the other thread with the example of AotS.
There will be times the 390x is going to be faster, and games where traditionally GCN was weaker the 480 has performance just above that of the 390x.
But as they say it is a partner card that may have slight/some OC.
Cheers

Supposedly all the cards out there have varying default clocks at the moment, which will be corrected by final BIOS or driver.

 

[CSI PC]

Supposedly all the cards out there have varying default clocks at the moment, which will be corrected by final BIOS or driver.

Most of that speculation came about from the guy who created that fake video and benchmarks.
Unfortunately he has messed up most information from a trust perspective, but according to WhyCry (who was adamant those other leaks-news were fake) the information he has is from an AMD partner.
Comes down to whether you trust WhyCry with his direct source information or not I guess.

Cheers

 

[Putas]

That should be 89 % for R9 390, not 79.

 

[CSI PC]

That should be 89 % for R9 390, not 79.

LOL shows how much most of us pay attention to the left colum, yeah pretty clear it is waaay out.
As you say the 480 is 11% faster than the 390.
Cheers